| 入江 琴子 | 1751013: M, 1回目発表 | ソフトウェア設計学 | 飯田 元☆ |
|
title: A Proposal for Taxonomy Toward the Impact Analysis on Software Behavior Resulted from Requirements Changes
abstract: It has been generally said that projects include 25% of requirements changes.To assure the high reliability of system,requests of requirements changes must be applied including their ripple effects. Change Impact Analysis(CIA) is the field that identifies the potential consequences of a change, or estimating what needs to be modified to accomplish a change, which is performed on requirements, architecture and software source code. In this presentation, I will introduce a paper "A software Change Impact Analysis Taxonomy", published in 2012. This paper proposed a taxonomy of the types of change impacts that can result from software source code, and used this taxonomy to classify change impact analysis algorithms based on the types of change impacts which a developer wanted to identify. However, the proposed taxonomy identifies only internal impacts on software. In practical CIA,the external impacts that are propagated by the behavior of the changed software source code must be identified, and then the decision whether to apply the change or not is required. At the end of this presentation, I will propose a taxonomy of the impacts from the viewpoint of things that propagate the impacts, which is expected to help the CIA on the level of a system's behavior. language of the presentation: Japanese 発表題目: 発表題目:要求の変更によるソフトウェアの振舞いへの影響分析に向けた分類の提案 発表概要: 平均的なプロジェクトでは要求の25%が変更されると言われている。システムやソフトウェアの高信頼性を保証するため、要求の変更はその影響も含めて正しく適用されなくてはならない。変更影響分析(Change Impact Analysis)は、変更が引き起こしうる結果や、変更を完了するために修正すべき場所を特定する分野で、要求や設計やソースコードの変更に対して行われる。 本発表では"A Software Change Impact Analysis Taxonomy"という2012年の論文を紹介する。この論文ではソースコードにおける変更影響の分類手法を提案し、さらに提案手法を変更影響分析アルゴリズムの分類に用いた。 しかし、既存手法のCIAはソフトウェア内部の影響箇所の識別にとどまっている。実際の影響分析においては影響箇所のソフトウェアの振る舞いがソフトウェアの外部に対してどのような影響を与えるのかを分析した上で修正の要否を検討する必要がある。そこで影響の媒介の視点で調査することで、振る舞いレベルへの影響を調査しやすくすることを提案する。 | |||
| 常木 健介 | 1751072: M, 1回目発表 | ソフトウェア設計学 | 飯田 元 |
|
title: コードクローンのタイプと保守作業実施の関係の調査
abstract:A code clone refers to a code fragment having the same or similar portion present in the source code. It is said that it is desirable to perform appropriate maintenance work (eg, refactoring tasks such as consistent remedial work and code aggregation) on code clones in order to prevent degradation of software maintainability. When many code clones are detected from large-scale software, it is difficult to perform maintenance work on everything. Furthermore, it is difficult to even find code clones that should be preferentially maintained. In existing research, for defect repair work at the time of bug occurrence, code clones are classified into three types and priority is investigated. However, in this research, we have not investigated maintenance work not related to defect repair.Therefore, in this research, we investigate the relationship between the type of code clone and maintenance work. This survey clarifies the maintenance work that should be applied preferentially for each type of code clone, and improvement of maintenance work efficiency can be expected. language of the presentation: Japanese | |||
| 中地 祥剛 | 1751077: M, 1回目発表 | ソフトウェア設計学 | 飯田 元 |
|
title: webアプリケーション開発における要求定義者のテスト記述支援システムの開発
abstract: In the web application development process, inconsistent communication of request specification between plannners and engineers frequently occurs. To prevent this problem, I will propose a system which supports planners can directly write test code using a simple DSL (domain description language). In this presentation, I will introduce several related research that writes test code from DSL and discuss the future direction of my research. language of presentation: Japanese | |||
| 中西 駿太 | 1751080: M, 1回目発表 | ソフトウェア設計学 | 飯田 元 |
|
title: Visualization of Discussions on Software Qualities in Open Source Software
abstract: To reduce software development cost, Open Source software (OSS) is frequently used in software development companies. However, using OSS with low software quality increase development cost. To prevent this problems, it is necessary to grasp the quality of target OSS in advance, however, it is difficult to grasp the quality of the OSS from available information. In this presentation, I present a system that that visualizes OSS developers’ discussions on software quality during decision-making and review processes. I hope that this system will be helpful for grasping quality information of OSS in software development companies . language of the presentation: Japanese 発表題目: OSS開発における品質に関する議論の可視化 発表概要: ソフトウェアの開発コストを削減するため、企業のソフトウェア開発現場においてOSS(オープンソースソフトウェア)の導入が盛んである。しかし、品質の低いOSSを導入すると開発コストの増加を招く恐れがある。これを防ぐため、導入の対象となるOSSの品質を事前に把握する必要があるが、OSSの品質を公開情報から把握することは困難である。本研究ではOSS開発における意思決定プロセスやレビュープロセスで行われた、OSS開発者による品質に関する議論を可視化するシステムを開発する。このシステムによって企業でOSS導入する際に品質情報の把握に役立つことが期待できる。 | |||
| 田口 勝弥 | 1751064: M, 1回目発表 | 自然言語処理学 | 松本 裕治☆ |
|
title: Novel Location De-identification for Machine and Human
abstract: Nowadays, the protection of personal information has drawn much attention, requiring an advanced technology on de-identification to remove personal information from data. When using data from Twitter, de-identification technique for Twitter should be developed. We assume two levels of location estimation; (1) a level of machine inferable location and (2)another level of human inferable location. To build the second-level estimation, we created a new corpus with a tag on human location inferable or not. By using the two types of corpora, we classify texts into several categories such as a machine-inferable but human-non-inferable tweet and so on. We also could obtain a de-identified tweet by iterations of removing the highest weighted word for a classifier. We believe our novel concepts on de-identification are essential for various privacy protection. language of the presentation: Japanese | |||
| 武田 悠佑 | 1751066: M, 1回目発表 | 自然言語処理学 | 松本 裕治☆ |
|
title: Robust Relational Data Clustering Considering Diversity of Noise
abstract: In this research, we propose robust relational data clustering considering diversity of noise. Analytical methods for relational data are regarded as important, while existing methods do not consider noises in real relational data , and they are not robust for the noises. We propose a method to eliminate various noise in relational data for clustering. language of the presentation: Japanese 発表題目: 多様なノイズに頑健な関係データクラスタリング 発表概要: 本研究では、ノイズの多様性を考慮した頑健な関係データクラスタリングを提案する．関係データの解析技術は重要視されているが，一方でこれまでの手法は実世界の関係データにおけるノイズが考慮されておらず，そのようなデータに対して頑健でないという問題がある．そこで，関係データにおける多様なノイズを除去した上でクラスタリングを行える手法を提案する． | |||
| 村山 太一 | 1751116: M, 1回目発表 | 自然言語処理学 | 松本 裕治☆ |
| title: propose and apply an indicator to capture deviation of word usage. abstract:In this presentation, I talk about new indicator to capture the deviation of words and proposed their application(simplification and normalization). First, using tweet data, we divide each word, and calculate deviation value using Gini coefficients of each word from various perspectives of person, time and position. Next, as one of the applications, we simplified the vocabulary by using the deviation value of a person and confirmed that the deviation value improves the accuracy. Finally, assuming that the deviation value of a person expresses the standard of a word rather than the word simplicity, we verified it using crowdsourcing. In the future, we will consider utilization for basic vocabulary selection and readability index, and how calculate deviation value for plural words. language of the presentation:japanese | |||
| 山本 英弥 | 1751121: M, 1回目発表 | 自然言語処理学 | 松本 裕治☆ |
|
title:Bidirectional Conversion Method between Inspection Values and Clinical Phenotypes in Medical Text abstract: The amount of medical texts has been increasing recently; it naturally leads to the situation of that secondary use of medical text data has drawn much attentions. However, there are a lot of problems in extracting medical information from the data written in natural language because natural language allows various ways of expression for one thing. For example, the concept of inflammation may be expressed by simply "inflammation" but it may also be expressed by using a pair of the inspection name and its value such as "CRP 25". To link a concept with its various expressions, this paper proposes a method to convert a pair of an inspection name and its value to a clinical phenotype. We used about 45,000 case reports collected from 2002 to 2015 as our materials. We extracted SIE (Sets of Inspection Expression) that consists of keyword, inspection value, and clinical phenotype. In order to evaluate the SIE, we used co-medicals' opinions and a medical guideline. As a result, SIE becomes similar to human understanding and do not match the guideline. However, we should be aware that SIE depends on who wrote an original medical text and what context diagnoses were taken. We suggest a method to deal with inspection values in medical texts and hope that the method triggers off developing a method for these expressions that machines cannot easily deal with. language of the presentation: Japanese | |||